Estimating the entropy of liquids from atom–atom radial distribution functions: silica, beryllium fluoride and water

Abstract

Molecular dynamics simulations of water, liquid beryllium fluoride and silica melt are used to study the accuracy with which the entropy of ionic and molecular liquids can be estimated from atom–atom radial distribution function data. The pair correlation entropy is demonstrated to be sufficiently accurate that the density–temperature regime of anomalous behaviour as well as the strength of the entropy anomaly can be predicted reliably for both ionic melts as well as different rigid-body pair potentials for water. Errors in the total thermodynamic entropy for ionic melts due to the pair correlation approximation are of the order of 10% or less for most state points, but can be significantly larger in the anomalous regime at very low temperatures. In the case of water, the rigid-body constraints result in larger errors in the pair correlation approximation, between 20 and 30%, for most state points. Comparison of the excess entropy, S e, of ionic melts with the pair correlation entropy, S 2, shows that the temperature dependence of S e is well described by T −2/5 scaling across both the normal and anomalous regimes, unlike in the case of S 2. The residual multiparticle entropy, ΔS = S e − S 2, shows a strong negative correlation with tetrahedral order in the anomalous regime.